Implementation of Q-learning algorithm for credit card fraud detection.

  • Bhavna Gopchandani et. al.


Financial fraud is an ever growing threat with way consequences within the economic business. Now-a-days the use of electronic transactions are increasing because they are easy and time saving for customers. Individual cards has its unique identification information which give services to the user and will be paid according to their use. German Credit Data is one of the most famous datasets in the realm of fraud detection and it is available in two formats that is categorical attribute and numerical attribute. Increased in number of credit card transactions opens the door for thieves to steal credit card details and commit fraud. Thieves steals the account of users which includes the card account number or other information that would be necessarily available to a merchant during legitimate transactions. To overcome this situation we are using Q- learning algorithm which is an  off policy reinforcement learning algorithm. Its considered off-policy because the Q- learning function learns from actions that are outside the current policy like taking random actions and therefore policy is not needed.  The Q in Q-learning algorithm stands for the quality that is applied on the raw and pre- processed knowledge ofdata.

This paper describes the existing application related to credit card fraud detection.